Routine Postoperative Antibiotic Prophylaxis Offers No Benefit after Hepatectomy—A Systematic Review and Meta-Analysis

Background: Prophylactic antibiotics are frequently administered after major abdominal surgery including hepatectomies aiming to prevent infective complications. Yet, excessive use of antibiotics increases resistance in bacteria. The aim of this systematic review and meta-analysis is to assess the efficacy of prophylactic antibiotics after hepatectomy (postoperative antibiotic prophylaxis, POA). Method: This systematic review and meta-analysis were completed according to the current PRISMA guidelines. The protocol has been registered prior to data extraction (PROSPERO registration Nr: CRD42021288510). MEDLINE, Web of Science and CENTRAL were searched for clinical reports on POA in hepatectomy restrictions. A random-effects model was used for synthesis. Methodological quality was assessed with RoB2 and ROBINS-I. GRADE was used for the quality of evidence assessment. Results: Nine comparative studies comprising 2987 patients were identified: six randomized controlled trials (RCTs) and three retrospectives. POA did not lead to a reduction in postoperative infective complications or have an effect on liver-specific complications—post-hepatectomy liver failure and biliary leaks. POA over four or more days was associated with increased rates of deep surgical site infections compared to short-term administration for up to two days (OR 1.54; 95% CI [1.17;2.03]; p = 0.03). Routine POA led to significantly higher MRSA incidence as a pathogen (p = 0.0073). Overall, the risk of bias in the studies was low and the quality of evidence moderate. Conclusion: Routine POA cannot be recommended after hepatectomy since it does not reduce postoperative infection or liver-specific complications but contributes to resistance in bacteria. Studies into individualized risk-adapted antibiotic prophylaxis strategies are needed to further optimize perioperative treatment in liver surgery.


Introduction
Surgical site infections (SSI) are the most prevalent complication after abdominal surgery, developing in up to 20% of patients [1]. In patients with SSI, mortality increases by over 12%, making strategies to reduce SSI a priority [1]. As the skin barrier is pierced during surgery, bacteria enter the abdominal cavity. In order to avoid clinically significant complications from this spread, standard hygiene measures must be adhered to [2]. In addition to

Study Selection
The study methodology was restricted to comparative studies. The design of comparative studies was unrestricted and included randomized-controlled trials (RCTs), propensity score-matched trials, prospective and retrospective reports. Comments, editorials, meeting abstracts, correspondence and reviews were excluded. The screening of titles, abstracts, as well as of full texts was carried out by two independent reviewers (AML and JF) and subsequently compared. All disagreements were resolved through discussion and consensus with consultation with another reviewer (KH).

Data Extraction
Data extraction was performed by two independent reviewers (AML and JF) using a standardized form prepared prior to extraction and adjusted based on the first three data extractions. Following raw data were extracted: title of the publication, year, author, country, journal, study design, number of study groups, the total number of patients, patient characteristics, factors, indications for hepatectomy, if and type of underlying liver disease was present, type of hepatectomy, rates of complications-SSI (overall, superficial and deep), remote infections, sepsis, post-hepatectomy liver failure (PHLF), bile leakage/bilioma, microbiota changes. The reviewers also noted the sources of funding for the included studies.

Statistical Analysis
Meta-analyses were performed using R (Version 4.0.3) packages "metafor", "meta", and "ggplot2". Summary statistics were performed using an independent t-test or χ 2test when applicable. Forest plots present effect estimates. A random-effects model was utilized for all outcomes. I 2 statistics were used to evaluate heterogeneity with an I 2 value below 25% indicating low, and over 75% indicating high heterogeneity. Odds ratios and 95% confidence intervals for dichotomous endpoints were pooled with the Mantel-Haenszel method.

Critical Appraisal of Included Studies
The risk of bias was assessed with Cochrane risk of bias tool 2 (RoB2) for RCTs and the ROBINS-I tool for non-randomized studies included. Certainty of evidence was assessed using GRADE. The assessments were performed by two independent reviewers (AML and JF) with consultation by a third reviewer on demand.

Results
After the exclusion of duplicates, 1876 records were screened for eligibility. After titles and abstract screening, 20 reports were assessed for eligibility based on full texts. Ultimately, nine reports were included in the qualitative and quantitative synthesis [12][13][14][15][16][17][18][19][20]. Figure 1 provides a detailed report of the study selection process.
Based on the results of the full-text screening, the following outcomes could be evaluated in the quantitative analysis: infective complications (overall surgical-site infections, superficial and deep SSIs, remote infections and sepsis) and liver-specific complications (PHLF and bile leakage). Table 1 provides details of included studies. Based on the results of the full-text screening, the following outcomes could be evaluated in the quantitative analysis: infective complications (overall surgical-site infections, superficial and deep SSIs, remote infections and sepsis) and liver-specific complications (PHLF and bile leakage). Table 1 provides details of included studies.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Gupta 2021 [8] Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Gupta 2021 [8] Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized control Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective c without randomization, the ROBINS-I tool was used (Table 3). Overa had a low risk of bias. Gupta 2021 [8] Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for ra Cochrane Risk of Bias tool, Version 2 ( Table 2). Fo without randomization, the ROBINS-I tool was use had a low risk of bias. Gupta 2021 [8] Low risk Some concern High risk.

Critical Appraisal of Included S
Risk of bias assessment was Cochrane Risk of Bias tool, Vers without randomization, the ROB had a low risk of bias.

Critical Apprai
Risk of bias a Cochrane Risk of without randomiz had a low risk of b

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using t Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative stud without randomization, the ROBINS-I tool was used (Table 3). Overall, included stud had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized control Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective c without randomization, the ROBINS-I tool was used (Table 3). Overa had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for ra Cochrane Risk of Bias tool, Version 2 ( Table 2). Fo without randomization, the ROBINS-I tool was use had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included S
Risk of bias assessment was Cochrane Risk of Bias tool, Vers without randomization, the ROB had a low risk of bias.

Critical Apprai
Risk of bias a Cochrane Risk of without randomiz had a low risk of b

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized control Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective c without randomization, the ROBINS-I tool was used (Table 3). Overa had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for ra Cochrane Risk of Bias tool, Version 2 ( Table 2). Fo without randomization, the ROBINS-I tool was use had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included S
Risk of bias assessment was Cochrane Risk of Bias tool, Vers without randomization, the ROB had a low risk of bias.

Critical Apprai
Risk of bias a Cochrane Risk of without randomiz had a low risk of b

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized control Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective c without randomization, the ROBINS-I tool was used (Table 3). Overa had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for ra Cochrane Risk of Bias tool, Version 2 ( Table 2). Fo without randomization, the ROBINS-I tool was use had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included S
Risk of bias assessment was Cochrane Risk of Bias tool, Vers without randomization, the ROB had a low risk of bias.

Critical Apprai
Risk of bias a Cochrane Risk of without randomiz had a low risk of b

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized control Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective c without randomization, the ROBINS-I tool was used (Table 3). Overa had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for ra Cochrane Risk of Bias tool, Version 2 ( Table 2). Fo without randomization, the ROBINS-I tool was use had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included S
Risk of bias assessment was Cochrane Risk of Bias tool, Vers without randomization, the ROB had a low risk of bias.

Critical Apprai
Risk of bias a Cochrane Risk of without randomiz had a low risk of b

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized control Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective c without randomization, the ROBINS-I tool was used (Table 3). Overa had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for ra Cochrane Risk of Bias tool, Version 2 ( Table 2). Fo without randomization, the ROBINS-I tool was use had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included S
Risk of bias assessment was Cochrane Risk of Bias tool, Vers without randomization, the ROB had a low risk of bias.

Critical Apprai
Risk of bias a Cochrane Risk of without randomiz had a low risk of b

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Low risk Some concern High risk.

Critical Appraisal of Included Studies
Risk of bias assessment was performed for randomized controlled trials using the Cochrane Risk of Bias tool, Version 2 ( Table 2). For retrospective comparative studies without randomization, the ROBINS-I tool was used (Table 3). Overall, included studies had a low risk of bias. Gupta 2021 [8] Low risk Some concern High risk.  Low , Moderate , High .

Patient Demographics
Cumulatively, 2987 patients were included in the meta-analysis. Most patients were male and had undergone an open hepatectomy for HCC. The summary demographics are provided in Table 4. Patients that did not receive post-operative antibiotics beyond POD1 (control) versus those patients that did (POA), did not vary significantly in age, gender, indications for surgery, type of surgery performed, or whether the surgery was performed through laparotomy or laparoscopically (Table 4).

Infective Complications
The main outcome of included studies and of this meta-analysis is surgical site infections. Seven studies [12][13][14][15][16]18,20] provided data for surgical site infection in patients receiving no postoperative antibiotics (control) compared to those given antibiotics until at least postoperative day 3 (POD3). Patients in the control group did not show higher rates of surgical site infections, whether deep, superficial, or overall, compared to those who received POA (Figure 2). Rates of overall remote infections did not differ significantly either and sepsis complications were comparable between groups (Figure 3).
A subgroup analysis of matched studies, as well as RCTs only was performed. Neither subgroup analysis revealed any significant effect of POA (Forest plots for subgroup analyses are provided as Supplementary Figures S1-S4).
Two studies [17,19] analyzed two regimens of prolonged POA-until POD2 (POA2) or for at least 4 PODs (POA4). Although there was no significant difference in both groups regarding overall and superficial surgical site infections, deep surgical site infections occurred significantly more frequently in patients receiving longer regimens (Figure 4).

Liver-Specific Complications
Liver-specific complications-PHLF or postoperative bile leakage-were reported in six studies [13][14][15][16]18,20]. Neither PHLF nor bile leakage occurred more frequently in either group ( Figure 5)  A random-effects model was utilized for all outcomes due to heterogenic methodological and clinical framework of included studies. Statistical heterogeneity was evaluated using the I 2 statistics. An I 2 value below 25% indicated low heterogeneity, while over 75% was considered high. A random-effects model was utilized for all outcomes due to heterogenic methodological and clinical framework of included studies. Statistical heterogeneity was evaluated using the I 2 statistics. An I 2 value below 25% indicated low heterogeneity, while over 75% was considered high. Rates of overall remote infections did not differ significantly either and sepsis complications were comparable between groups ( Figure 3). A random-effects model was utilized for all outcomes due to heterogenic methodological and clinical framework of included studies. Statistical heterogeneity was evaluated using the I 2 statistics. An I 2 value below 25% indicated low heterogeneity, while over 75% was considered high.
A subgroup analysis of matched studies, as well as RCTs only was performed. Neither subgroup analysis revealed any significant effect of POA (Forest plots for subgroup analyses are provided as Supplementary Figures S1-S4).
Two studies [17,19] analyzed two regimens of prolonged POA-until POD2 (POA2) or for at least 4 PODs (POA4). Although there was no significant difference in both groups regarding overall and superficial surgical site infections, deep surgical site infections occurred significantly more frequently in patients receiving longer regimens (Figure 4). A random-effects model was utilized for all outcomes due to heterogenic methodological and clinical framework of included studies. Statistical heterogeneity was evaluated using the I 2 statistics. An I 2 value below 25% indicated low heterogeneity, while over 75% was considered high.
Methicillin-resistant Staphylococcus aureus (MRSA) was detected significantly more as the pathogenic isolate in patients with routine POA, while S. aureus (without resistance) was detected significantly more often in the control group (p = 0.007293 and p = 0.000399, respectively, as determined by the χ 2 -test). The incidence of S. aureus as the pathogenic isolate regardless of the resistance was similar in both groups (p = 0.507111). All other pathogens did not show a statistically significant variation in this analysis. The association between deep SSIs and MRSA could not be tested due to insufficient raw data provided by individual studies.

Certainty of Evidence
GRADE (grading of recommendations assessment, development and evaluation) approach was used to rate the certainty of evidence. An overview of the main outcomes is provided in Table 5.

Liver-Specific Complications
Liver-specific complications-PHLF or postoperative bile leakage-were reporte six studies [13][14][15][16]18,20]. Neither PHLF nor bile leakage occurred more frequently in ther group ( Figure 5). A random-effects model was utilized for all outcomes due to heterogenic methodological and clinical framework of included studies. Statistical heterogeneity was evaluated using the I 2 statistics. An I 2 value below 25% indicated low heterogeneity, while over 75% was considered high. A random-effects model was utilized for all outcomes due to heterogenic methodological and clinical framework of included studies. Statistical heterogeneity was evaluated using the I 2 statistics. An I 2 value below 25% indicated low heterogeneity, while over 75% was considered high.
Subgroup analysis for PHLF did not reveal any significant difference (

Pathogenic Isolates
The most common pathogenic isolates for control versus POA were provided by six studies comparing POA versus control [12,[14][15][16]18,20] and are depicted in Figure 6. A random-effects model was utilized for all outcomes due to heterogenic methodological and clinical framework of included studies. Statistical heterogeneity was evaluated using the I 2 statistics. An I 2 value below 25% indicated low heterogeneity, while over 75% was considered high. Methicillin-resistant Staphylococcus aureus (MRSA) was detected significantly more as the pathogenic isolate in patients with routine POA, while S. aureus (without resistance) was detected significantly more often in the control group (p = 0.007293 and p = 0.000399, respectively, as determined by the χ 2 -test). The incidence of S. aureus as the pathogenic isolate regardless of the resistance was similar in both groups (p = 0.507111). All other pathogens did not show a statistically significant variation in this analysis. The association between deep SSIs and MRSA could not be tested due to insufficient raw data provided  Overall, the quality of evidence was rated as moderate with two outcomes (PHLF and sepsis) downgraded due to the low number of studies.

Discussion
Liver surgery is a rapidly growing surgical specialty with expanding indications and novel techniques appearing and constantly evolving [21,22]. In addition to technique optimization, it is important to improve perioperative management. With complication rates reaching 31% for hepatectomy, validated strategies to reduce morbidity are wanted [23]. Postoperative antibiotic prophylaxis has been a subject of debate recently as more and more studies failed to show the benefit of this practice even in cases of intraoperative contamination [24][25][26]. There is an exceeding need for improving strategies for antibiotic use within clinical settings. Previous network meta-analysis of five studies has compared various antibiotic strategies and did not reveal a benefit of additional antibiotics, findings which were confirmed in the current meta-analysis of nine studies [27]. The growth and spread of multidrug-resistant bacteria lead to increased mortality, morbidity and healthcare cost [6]. A major driver of multidrug resistance is sporadic and unnecessary antibiotic use. An evidence-based approach to antibiotic administration is needed and with this, common, empirically-driven practices must be evaluated and verified. Current meta-analysis suggests an increase of resistance in pathogens after postoperative antibiotic prophylaxis without the benefit of a reduced rate of overall infections.
After hepatectomy, liver function suffers from a transient decline, which is predominantly evident by systematic measurements of protein synthesis and coagulation parameters [8]. How hepatic immunologic function is affected is less clear. Since antibiotics are predominantly metabolized by the hepatocytes, strategies to reduce potential injury to the liver are invaluable. During the perilous phase where the liver must regenerate to accommodate the demands of the organism, the question emerges if antibiotics provide additional support or a hurdle.
This meta-analysis provides first glimpses into the efficacy of postoperative prophylactic antibiotics after hepatectomy. Prophylactic use of postoperative antibiotics fails to show any significant improvement in terms of infective or liver-specific complications. In fact, prolonged administration of antibiotics for more than four days was associated with more deep surgical site infections. Although counterintuitive at first glance, antibiotics only select bacteria, which leads to increased resistance [6]. With more resistance, difficult-to-treat bacteria can spread and become problematic for the organism to clear. With drug-resistant bacteria becoming a major burden on healthcare due to high associated mortality, morbidity and treatment cost, reduction of resistance becomes an increasingly important research area [28]. Postoperative antibiotic prophylaxis after hepatectomy leads to increased multidrug resistance in Staphylococcus aureus. Whether other resistances increase as well, e.g., in bacteria that were not primarily pathogenic, remains unclear. Long-term consequences of increased resistant bacteria, in particular MRSA, in patients after hepatectomy, too, remain unclear but can only be anticipated as undesirable.
Increasing evidence suggests that when it comes to antibiotics, less is more unless it's sepsis. With this thought in mind, clinicians must critically reexamine antibiotic strategies. The appraisal for postoperative antibiotic prophylaxis after hepatectomy reveals the strategy as inadequate and superfluous, thus one that should be abandoned.

Certainty of Evidence
The main limitation of the analysis is the limited number of studies on the topic. Simultaneously, with most RCTs, the quality of evidence is satisfactory. Various studies had different regimens implemented with most comparing no post-operative antibiotics versus three days. However, three doses versus nine, as well as zero versus seven days of antibiotics were also compared. This brings a certain heterogeneity into the analysis. Different antibiotics were utilized as prophylactics with principally flomoxef sodium administered.
The risk of bias in included studies was generally low and the certainty of the evidence was overall moderate based on the methodology of the studies. This is the highest certainty of evidence available to date.

Conclusions
Routine postoperative antibiotic prophylaxis cannot be recommended for patients after hepatectomy and the administration should be discouraged. Prolonged exposure to antibiotic prophylaxis leads to an increase in resistant bacteria and thus bears an additional risk for patients. Prophylactic antibiotic strategies must be assessed for their efficacy and safety to further optimize liver surgery. Studies into individualized risk-adapted antibiotic prophylaxis strategies are needed to further optimize perioperative treatment in liver surgery.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/antibiotics11050649/s1, Table S1: Search strategy; Figure S1: Forest plot for overall surgical site infections in patients with POA versus control-matched subgroup; Figure S2: Forest plot for overall remote infections in patients with POA versus control-matched subgroup; Figure S3: Forest plot for overall surgical site infections in patients with POA versus control-RCT subgroup; Figure S4: Forest plot for overall remote infections in patients with POA versus control-RCT subgroup. Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: All data generated or analyzed during this study are included in this published article and its Supplementary Information Files.